BaMM

BaMM performs de-novo motif discovery and regulatory sequence analysis using higher-order Bayesian Markov Models (BaMMs) to model transcription factor binding specificity.

Key Features:

• De-novo motif discovery: identifies enriched sequence motifs within input nucleotide sets using BaMMs.
• Bayesian Markov Models (order 4): represents motifs with fourth-order Bayesian Markov Models to capture higher-order nucleotide dependencies.
• Improved predictive performance: shows superior receiver operating characteristic (ROC) performance compared to position weight matrices (PWMs) and first-order models.
• AvRec scoring: quantifies motif quality using the AvRec score, defined as average recall over the true positive-to-false positive ratio between 1 and 100.
• Motif scanning: scans nucleotide sequences with pre-identified motifs to locate motif occurrences.
• Motif similarity search: searches for motifs similar to a query motif within the motif database to identify related transcription factor motifs.
• Motif database trained on GTRD ChIP-seq: includes motifs trained using GTRD ChIP-seq data and contains motifs associated with over 1000 transcription factors.

Scientific Applications:

• Regulatory element identification: discovery of enriched sequence motifs in genomics and transcriptomics datasets.
• Regulatory element localization: localization of motif occurrences that may influence gene expression through motif scanning.
• Motif annotation and TF assignment: identification of related transcription factor motifs via motif similarity search against a GTRD-trained database.
• Gene regulation and TF binding studies: support for analyses of transcription factor binding specificity and regulatory network investigation.

Methodology:

Represents motifs as fourth-order Bayesian Markov Models, trains motifs on GTRD ChIP-seq data, evaluates performance with ROC analyses and the AvRec score (average recall over TP/FP ratio 1–100), and performs motif scanning and motif similarity searches.